Garrison Lovely thinks resistance to AI that replaces human work can become a mass movement. He also thinks it can fail by picking the wrong target. Lovely is a journalist and the author of the book Obsolete. He now sits on the nonprofit board of Irreplaceable, a group organizing against what he calls the race to replace us. On The Cognitive Revolution, in an episode published on 29 September 2026, he explained how he thinks that opposition could gain political force. His plan borrows from climate organizing and treats local data-center fights as a starting point, not the end goal. He also said AI has so far escaped the partisan split that defines most American issues.
Why he joined Irreplaceable
Lovely said that as he finished the book he concluded that "we need a mass movement organized to, you know, stop the race to replace us." He agreed with some of what the existing groups had done and disagreed with other parts. The gap he saw was for organizing that frames AI "not just in terms of risk, but also democracy," led by people who have built movements before. When Irreplaceable appeared, he said, it filled "exactly that gap." The group asked him to join its board, and he is donating his share of the book's royalties to it.
In the episode, the group's aims were summed up in a line from its website: winning "a say, a stake and a slowdown." Irreplaceable's platform sets out three demands behind that phrase:
- public authority over where and how AI is used
- an economy that spreads AI's benefits beyond the executives and owners of AI companies
- preventing systems that pose existential risks, meaning risks to humanity's survival
The climate movement's lesson
Lovely said many of the people who started and lead Irreplaceable came from the climate movement, which he said "gets a bad rap in some ways." Even so, he argued, it took an issue that was not a political winner, turned it into a real force, and won results such as the Inflation Reduction Act, the 2022 US law that included large clean-energy measures.
He passed on a story from Irreplaceable's director, Phil Aroneanu, who co-founded the climate group 350.org with Bill McKibben and others. According to Lovely, Aroneanu described a split among climate activists a decade or two ago. One camp focused on immediate harms such as pollution, oil spills and coal plants. The other focused on greenhouse-gas emissions. Lovely compared this to today's AI debate between people focused on present harms and people focused on existential risk. People worried about existential risk sometimes argue that nobody would call cleaning up an oil spill a distraction from climate change. Lovely's point was that climate activists really did have that fight, and then "managed to figure it out and bury the hatchet and then work together."
Aroneanu's founder's note draws on the same background. He argues that concentrated corporate influence has to be met with organized public participation, not persuasion alone. The work he describes runs from school and county meetings up to state legislatures and national demonstrations.
For AI, Lovely wants a "big tent" coalition with clear demands built around the message that people do not want to be replaced by machines. It has to reach beyond the in-group, he said, "because this is really going to take a lot of people to effectively resist the wealthiest industry of all time."
'Yes, and' on data centers
Data centers are the large buildings full of computer chips that train and run AI systems. Local fights over them were the most concrete test of this strategy in the conversation. The discussion raised a caution that restricting data centers might backfire: the price of renting GPU time (the specialized chips AI runs on) is already rising, so limits could price ordinary people out of everyday AI use. It also included wariness about requiring majority approval when a landowner wants to build.
Lovely said people who want to stop AI sometimes treat local opposition as proof that the public is with them, but the polling is "more complicated." Much of the opposition, he said, is about local environmental or community effects. Some of those concerns are legitimate. Others he considers "quite overstated," the product of "a lot of bad reporting." He called the inequality concern worth taking seriously, but said the bigger issue is the risk the technology poses if it keeps developing as it is.
His answer is a "yes and" approach. He would welcome someone fighting a local data center and then ask whether they worry about AI's effect on society. If they do, he would tell them that blocking one project will not change much. Even a national moratorium, he warned, would disappoint people who expect it to meaningfully slow AI. He gave three reasons:
- Many projects are already underway.
- Existing data centers can swap in newer chips.
- AI is already helping to automate parts of its own development.
He noted that AI progress has not hit a wall since 2012. Without regulation, he expects AI to move faster in the future, and he worries that campaigners who win a hard-fought moratorium will then ask why things keep "getting crazier."
He said a moratorium could work as part of a broader package. His example was the proposal from Senator Bernie Sanders and Representative Alexandria Ocasio-Cortez. Lovely said it started as a simple moratorium and later gained limits on exports of advanced AI chips. Under those limits, no country could receive the chips without strong AI safety rules and conditions on green energy and union labor. The sponsors' announcement of the AI Data Center Moratorium Act, dated 25 March 2026, presents the pause and the export restrictions together. It ties lifting the pause to safeguards on safety, privacy and civil rights, workers' share of the gains, utility costs, communities and the environment. Lovely called it a messaging bill that is "probably not going to pass," but one that now looks "quite prescient" as the country has turned against data centers. He said it would meaningfully slow progress compared with where it would otherwise be.
Still, he argued that the real leverage lies elsewhere. "We need to regulate the model developers and then the chip makers," he said, calling that "the only way" to change how the technology arrives.
A rare unpolarized issue
On party politics, Lovely called it a "remarkable feature" that AI has long been one of the only US issues where concern and support for regulation are similar across parties. He said there was very recent evidence of some polarization. Republican concern dropped after President Trump came out hard against AI risk and called it a hoax, but Lovely said it was "not a huge drop." He also pointed to Republicans who sound worried. He cited Vice President JD Vance saying that if the companies are building Frankenstein, they should stop. He said Kratsios, who heads the White House Office of Science and Technology Policy, made a similar remark about the companies being able to stop.
Lovely recalled that Trump had disparaged communities that do not want data centers, which he remembered roughly as calling them "stupid and poor or something." He does not think that will polarize the issue. Once an issue becomes this prominent, he argued, people hold "their own deeply felt convictions" about it. He listed their worries: jobs, the environment, the concentration of power and wealth, and surveillance. He placed the backlash against Flock surveillance cameras in that same wider resistance to what he sees as a dystopian tech future.
He does not think advocates can prevent polarization by staying quiet. Senator Elizabeth Warren has supported a pause, and he said nobody will stop people on the left from saying such things. A popular position embraced by one side need not split along party lines, he argued, and "it could just make people think better of them." When Sanders came out against Flock, Lovely said, many people quote-tweeted him to say, in effect, that they disliked Sanders but agreed with him on this.
Lovely described a "societal immune response" that is arriving ahead of schedule, though he added that the risks are arriving ahead of schedule too. He said he would feel better with "basically anybody else" in the White House, because Trump is "uniquely insulated from what the public wants." He compared Trump's current stance to COVID in February 2020: early denial, then eventually taking it more seriously. If the harms become "abundantly clear," as Lovely expects, he thinks Trump will struggle to hold his position.
From journalist to advocate
Lovely said he began Obsolete as a journalist documenting events and making an argument, and felt uncomfortable making policy recommendations. His interviews changed that. He came to believe that if AI keeps getting more capable, the default outcome is "doom or dystopia," and that which one we get depends on whether AI systems do as they are told, which he called "currently an open question."
He said he still believes in the potential of deep learning, the technique behind today's AI systems. But he thinks it is "being pointed to the wrong things by the wrong people for the wrong reasons." The fix, in his view, is democratic governance of the technology, "small D democratic," as he put it. His message to people inside AI companies was that they will be replaced and lose their power too. He said many insiders, and even the CEOs, share the worry but feel stuck in a prisoner's dilemma, where no one wants to hold back while rivals race ahead. In his view, "the government and the companies, they can't be trusted to do the right thing here," so the solution lies with the public, which has to "make it easy for them."
He is also starting a podcast, Organize Against the Machine, with labor organizer Cassie Pritchard. The aim is to turn the book's ideas into real-world action.